Commits · 855cb1486af3a9e7ee91a6b3038323c291ecf87b · OpenDAS / vllm_cscc

16 Jan, 2026 2 commits
- update unified_attention_with_output_fake · 9d16d5aa
  zhuwenwen authored Jan 16, 2026
  
  9d16d5aa
- add VLLM_USE_FUSED_CACHE_QUANT_BMM_MLA to use fused rmsnorm + contiguous +... · 9dd70f0e
  zhuwenwen authored Jan 16, 2026
```
add VLLM_USE_FUSED_CACHE_QUANT_BMM_MLA to use fused rmsnorm + contiguous + rope(for dpsk-v3) + concat_and_cache_mla + q quant, control bmm(todo) + cat +mla (fp8)
```
  9dd70f0e
23 Dec, 2025 1 commit
- add VLLM_USE_FUSED_FILL_RMS_CAT for dpsk mtp fill + rms*2 + cat · e80dcabe
  zhuwenwen authored Dec 23, 2025
```
update VLLM_USE_LIGHTOP_RMS_ROPE_CONCAT impl
```
  e80dcabe
18 Dec, 2025 2 commits
- update self._q_scale · ee93cb70
  zhuwenwen authored Dec 18, 2025
  
  ee93cb70
- update get_mla_decoding_metadata_dense_fp8 interface and _k_scale&_v_scale · 1acf2d7a
  zhuwenwen authored Dec 18, 2025
  
  1acf2d7a
17 Dec, 2025 2 commits
- convert q to float8_e4m3fn · da85feb7
  zhuwenwen authored Dec 17, 2025
  
  da85feb7
- remove fuse_rmsnorm_rope_quant_gfx938 · 99981972
  zhuwenwen authored Dec 17, 2025
  
  99981972
16 Dec, 2025 1 commit
- add fuse_rmsnorm_rope_quant_gfx938 to support use fp8_e4m3 mla · 0ce3b670
  zhuwenwen authored Dec 16, 2025
  
  0ce3b670
01 Dec, 2025 1 commit
- add VLLM_USE_OPT_RESHAPE_AND_CACHE (test) · 64e307c7
  zhuwenwen authored Dec 01, 2025
  
  64e307c7
19 Nov, 2025 1 commit
- set VLLM_USE_LIGHTOP_FILL_MOE_ALIGN=0 · 46006aee
  zhuwenwen authored Nov 19, 2025
  
  46006aee
13 Nov, 2025 2 commits
- [fix]解决moe_fused_gate编译错误，去掉mla中mtp部分的修改 · b956fc64
  zhuwenwen authored Nov 13, 2025
```
restore the default settings of disable_cascade_attn
add VLLM_USE_OPT_ZEROS to replace triton_ (torch.zeros)
set default_max_num_batched_tokens = 10240
update qwen3_moe of layernorm
```
  b956fc64
- add VLLM_USE_OPT_ZEROS to replace triton_ (torch.zeros) · b364c176
  zhuwenwen authored Nov 13, 2025
  
  b364c176
12 Nov, 2025 1 commit
- update qwen3_moe of layernorm and activation · a014d6a5
  zhuwenwen authored Nov 12, 2025
  
  a014d6a5
29 Oct, 2025 2 commits
- remove fp8_e5m2 error · 397b9edd
  zhuwenwen authored Oct 29, 2025
  
  397b9edd
- remove fp8_e5m2 error · 2b90ecd3
  zhuwenwen authored Oct 29, 2025
  
  2b90ecd3
14 Oct, 2025 1 commit
- remove redundant maybe_calc_kv_scales · c3b8a0ae
  zhuwenwen authored Oct 14, 2025
  
  c3b8a0ae
13 Oct, 2025 1 commit
- 新增fp8—e5m2 · ed3cdc81
  zhuwenwen authored Oct 13, 2025
  
  ed3cdc81
10 Oct, 2025 3 commits
- keep same version · fc5bfc66
  maxiao1 authored Oct 10, 2025
  
  fc5bfc66
- pd分离_tbo · 12291212
  maxiao1 authored Oct 06, 2025
  
  12291212
- When VLLM_USE_FLASH_ATTN_PA=0, support the use of default block_size · 3daae57c
  zhuwenwen authored Oct 10, 2025
  
  3daae57c
09 Oct, 2025 1 commit
- Revert "Merge branch 'v0.9.2-dev-lmcache-pd' into 'v0.9.2-dev'" · 7b30ecd8
  zhuwenwen authored Oct 09, 2025
```
This reverts merge request !219
```
  7b30ecd8
28 Sep, 2025 1 commit
- 修复kvcache-fp8—e5m2的不能开cp的bug · b2d14ba3
  yangql authored Sep 28, 2025
  
  b2d14ba3
24 Sep, 2025 2 commits
- add env · c6b7a44b
  yangshj1 authored Sep 24, 2025
  
  c6b7a44b
- add engine id · f6fcc8ff
  xuxz authored Sep 24, 2025
  
  f6fcc8ff
23 Sep, 2025 1 commit
- 修复kvcachefp8与cp/pc的冲突 · 8b7daa0d
  yangql authored Sep 23, 2025
  
  8b7daa0d
17 Sep, 2025 1 commit
- fix the performance issue of tbo pd separation · f8985b96
  lizhigong authored Sep 17, 2025
  
  f8985b96
16 Sep, 2025 1 commit
- add pd lmcache · f4fd3a96
  yangshj1 authored Sep 16, 2025
  
  f4fd3a96
19 Aug, 2025 1 commit
- update v1 fa layout · 864c718a
  zhuwenwen authored Aug 19, 2025
  
  864c718a
06 Aug, 2025 1 commit
- update VLLM_FLASH_ATTN_V1 to VLLM_USE_FLASH_ATTN_PA · 88dbf92c
  zhuwenwen authored Aug 06, 2025
  
  88dbf92c
05 Aug, 2025 1 commit
- when using VLLM_FLASH_ATTN_V1, set block_size to 64 · 80a682c7
  zhuwenwen authored Aug 05, 2025
  
  80a682c7
17 Jul, 2025 2 commits
- update adpaptive block_size · 937a3ec1
  zhuwenwen authored Jul 17, 2025
  
  937a3ec1
- set default block_size and pa · b61f7a69
  zhuwenwen authored Jul 17, 2025
  
  b61f7a69
06 Jul, 2025 1 commit
- [V1] Support any head size for FlexAttention backend (#20467) · 9fb52e52
  Cyrus Leung authored Jul 07, 2025
```
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
```
  9fb52e52
26 Jun, 2025 1 commit

[Bugfix][V1][ROCm] Fix AITER Flash Attention Backend (Fix API Break and Local... · 27c065df

TJian authored Jun 26, 2025


[Bugfix][V1][ROCm] Fix AITER Flash Attention Backend (Fix API Break and Local Attention Logic: affecting Llama4) (#19904)
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com>

27c065df

18 Jun, 2025 1 commit
- [Minor] Zero-initialize attn output buffer (#19784) · f04d6045
  Woosuk Kwon authored Jun 17, 2025
```
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
```
  f04d6045
15 Jun, 2025 1 commit
- [Kernel] Raise verbose error and consolidate `num_heads/num_kv_heads` divisibility check (#19339) · 0b73736a
  22quinn authored Jun 14, 2025
```
Signed-off-by: 22quinn <33176974+22quinn@users.noreply.github.com>
```
  0b73736a
12 Jun, 2025 1 commit
- [torch.compile][ROCm] Fuse quantization onto attention using a torch.compile pass (#16756) · f98548b9
  Luka Govedič authored Jun 12, 2025
```
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
```
  f98548b9
03 Jun, 2025 2 commits
- [V1] Support cross-layer KV sharing (#18212) · bdf13965
  Yong Hoon Shin authored Jun 03, 2025
```
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
```
  bdf13965
- [Misc] Add SPDX-FileCopyrightText (#19100) · 02f0c7b2
  Simon Mo authored Jun 03, 2025
```
Signed-off-by: simon-mo <simon.mo@hey.com>
```
  02f0c7b2
06 May, 2025 1 commit
- [v1] AttentionMetadata for each layer (#17394) · cba31c47
  Chen Zhang authored May 06, 2025
```
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
```
  cba31c47